How to construct normal reference values using spss?

[jabs]

I have been trying to construct normal reference values for a
biometric fetal parameter but after finishing this process, the
equations for the median and the standard deviation do not seem to fit
with the values when I run these equations to look for the normal
distribution.
The statistical method to estimate reference intervals was the one
described by Royston and Wright.
Do you know (step by step) how to construct normal reference values
using spss?
Thanks for your help !!!

[thomas]

First, I recomment software Medcalc (www.medcalc.be), it calculates
reference intervals using parametric and non-parametric methods.
Furthermore, it calculates CONFIDENCE INTERVALS of estimates for
reference limits.

You get meaningful reference values, if you might have a sample size
of at least 120 subjects (see CLSI guideline C28-4).

In SPSS, you could calculate 2.5% + 97.5% percentiles,or other
percentiles (95%), using -->analyze --> frequency --> statistics.

FREQUENCIES
VARIABLES=M4.1 /FORMAT=NOTABLE
/PERCENTILES= 95
/ORDER= ANALYSIS .

However, there are a lot of possibilities how to estimate a percentile
with SPSS (Default is waverage, ...).
Search for "Methods for computing percentiles" within Google, and look
at the MedStats-entry. Here, several options how to calculate
percentiles are discussed.

And: the CI of reference limits are not given in SPSS.

The free available paper describes calculation of CI:
Campbell MJ, Gardner MJ (1988): Calculating confidence intervals for
some non-parametric analyses. British Medical Journal, 296, 1454-1456
http://www.pubmedcentral.nih.gov/picrender.fcgi?artid=1833334&blobtype=pdf

Best regards
Thomas Keller
ACOMED statistik

[m***@student.rug.nl]

Tha ta poume sto spiti

[Kuenzang Dorji]

Post by m***@student.rug.nl
Tha ta poume sto spiti

Can I know the name of the article? the link is not working. Thanks
--
This message and/or any associated files may contain confidential
and/or privileged information intended solely for the use of the individual
or entity to whom they are addressed. If you are not the intended or
authorized recipient, you must not use, copy, disclose or take any action
based on this message or any information contained therein. If you have
received this message in error, please advise the sender immediately by
sending a reply email and delete this message. The recipient should check
this email and any attachments for the presence of viruses. The Royal
Government of Bhutan accepts no liability for any damage caused by any
virus transmitted by this email or any/or associated files.

[Rich Ulrich]

On Wed, 6 Oct 2021 00:06:57 -0700 (PDT), Kuenzang Dorji

Post by Kuenzang Dorji

Post by m***@student.rug.nl
Tha ta poume sto spiti

Can I know the name of the article? the link is not working. Thanks

Here is the message starting the topic, which you failed to
include with the question. I found this by looking at the
"references" in your header. It is not very helpful.
= = = start
I have been trying to construct normal reference values for a
biometric fetal parameter but after finishing this process, the
equations for the median and the standard deviation do not seem to fit
with the values when I run these equations to look for the normal
distribution.
The statistical method to estimate reference intervals was the one
described by Royston and Wright.
Do you know (step by step) how to construct normal reference values
using spss?
Thanks for your help !!!
= = = end

When I Google on most of the Subject, I also find this
further (original) response, which may be the article
you ask about, and which you should have quoted with
your question.

= = = start
First, I recomment software Medcalc (www.medcalc.be), it calculates
reference intervals using parametric and non-parametric methods.
Furthermore, it calculates CONFIDENCE INTERVALS of estimates for
reference limits.

You get meaningful reference values, if you might have a sample size
of at least 120 subjects (see CLSI guideline C28-4).

In SPSS, you could calculate 2.5% + 97.5% percentiles,or other
percentiles (95%), using -->analyze --> frequency --> statistics.

FREQUENCIES
VARIABLES=M4.1 /FORMAT=NOTABLE
/PERCENTILES= 95
/ORDER= ANALYSIS .

However, there are a lot of possibilities how to estimate a percentile
with SPSS (Default is waverage, ...).
Search for "Methods for computing percentiles" within Google, and look
at the MedStats-entry. Here, several options how to calculate
percentiles are discussed.

And: the CI of reference limits are not given in SPSS.

The free available paper describes calculation of CI:
Campbell MJ, Gardner MJ (1988): Calculating confidence intervals for
some non-parametric analyses. British Medical Journal, 296, 1454-1456
http://www.pubmedcentral.nih.gov/picrender.fcgi?artid=1833334&blobtype=pdf

Best regards
Thomas Keller
ACOMED statistik
= = = end

Yeah, that nih link did not work. Shortening it did not help (which
sometimes does help).

When I Googled on much of the title of the paper, I found
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2545906/

- which seems to be the right reference.

Google can be your friend.